Please enable JavaScript.
Coggle requires JavaScript to display documents.
Leverage machine learning pipelines to explore ventricular features in…
Leverage machine learning pipelines to explore ventricular features in large clinical ultrasound data sets
case: Second highest mortality rate, heart disease (elderly) with 3 years treatment
data: cardiac ultrasound (1202 medical record with 36994 pieces) from 403 patients
9 groups: group11,group12,group13,group23,group22, group 31, group 32, group 33
3 categories treatment method : cardiac catheterization, ventricular defibrillator, drug control
Majority G23 (verticular defibrillator and follow-up)
Feature selection: RFE (from 20 variable)
identification patients based on ultrasound data using machine learning algorithms
Supervised
Random Forest
XGBoost
Light Gradient Boosting Machine
Unsupervised
Voting Ensemble
Gaussian Mixture Clustering
K-Means Algorithm
Mean Shift Algorithm
Feature Selection: Recurcive Feature Elimination (RFE) algorithm, regression method
age, Left Ventricular (LV), Ventricular Septum (VS), Left Ventricle Posterior Wall (LVPW), Left Atrium (LA), Aorta (AO), TR Mean, Pressure Gradient (PG), Left Ventricular Ejection Fraction (LVEF)
Evaluation
Silhouette coefficient
Calinski_harabasz evaluation
Davies-Bouldin Index
Euclidean distance
Tools: Auto ML, PyCaret library